Hide addresses in traced code.

Hide addresses in traced code.

BUG=

[Sven Panne, 2013-01-21 09:51:05]

Quick DBC. From a layering perspective this CL is ugly, because the client of
the string stream should handle this, not the stream itself, but I can see the
point of having a local (wrong) change...

https://chromiumcodereview.appspot.com/12033007/diff/1/src/string-stream.cc
File src/string-stream.cc (right):

https://chromiumcodereview.appspot.com/12033007/diff/1/src/string-stream.cc#n...
src/string-stream.cc:171: void* value = current.data_.u_pointer_;
I would prefer just a ternary ?: to calculate value (e.g. NULL when we don't
want to see the actual pointer) instead of the whole if/else.

https://chromiumcodereview.appspot.com/12033007/diff/1/src/string-stream.cc#n...
src/string-stream.cc:213: if (FLAG_trace_code_addresses) {
Do we really need this if we have the change above? It looks like this is
handled there already in an identical way.

[Massi, 2013-01-21 15:27:49]

> From a layering perspective this CL is ugly, because the
> client of the string stream should handle this, not the
> stream itself, but I can see the point of having a local
> (wrong) change...

Yes, I know this is a bit ugly but I would find patching the client even uglier
in practice.
After all this is just a debugging aid.

I renamed the flag to state what it does even more clearly. 

We could also decide not to commit it, it is not *strictly* needed, it is just
very convenient (at least to me).
And since by default it's off it should not hurt anybody :-)

https://chromiumcodereview.appspot.com/12033007/diff/1/src/string-stream.cc
File src/string-stream.cc (right):

https://chromiumcodereview.appspot.com/12033007/diff/1/src/string-stream.cc#n...
src/string-stream.cc:171: void* value = current.data_.u_pointer_;
On 2013/01/21 09:51:05, Sven Panne wrote:
> I would prefer just a ternary ?: to calculate value (e.g. NULL when we don't
> want to see the actual pointer) instead of the whole if/else.

The code would be more "linear" but the output would display "NULL pointers"
instead of "hidden values".
I would prefer to have the explicit "0xXXXXXXXX" string in the log so that
nobody will ever mistake it for a NULL.

https://chromiumcodereview.appspot.com/12033007/diff/1/src/string-stream.cc#n...
src/string-stream.cc:213: if (FLAG_trace_code_addresses) {
On 2013/01/21 09:51:05, Sven Panne wrote:
> Do we really need this if we have the change above? It looks like this is
> handled there already in an identical way.

Done.

[Sven Panne, 2013-01-22 07:33:40]

The reason for this CL is probably making the comparison of different
hydrogen.cfgs easier, isn't it? If this is really the only reason, I would
object to this CL, because a trivial sed script can achieve basically the same
without adding yet another cruel hack to v8:

   sed 's/\<0x[[:xdigit:]]*\>/0xXXXXXXXX/g'

[Michael Starzinger, 2013-01-24 13:45:26]

I agree with Sven, I think it is better to do that in a post-processing step.